Acceptance testing was one of the first challenges I faced during my apprenticeship. I learned that complexity scales in proportion to the number of scenarios, configurations and components being tested.
In my opinion, there is no one solution to solve all our problems. Having said that, I have noticed a few patterns that tend to stall the proceeding of acceptance testing practices. Incidentally, these patterns are largely identical to the rules from Clean Test section in Clean Code (132).
The first rule: fast
Acceptance testing automates the process of walking through the application interface and checks for intended behaviors. Unlike unit testing, which drills down to a very specific function, testing for behaviors requires launching several components, and sometimes the entire system in order to get a result.
To save time, we schedule our test suites to run on a separate machine. We integrate with the Hudson CI server, in particular, for its friendly user interface and various plugins that support the scripting tools and testing frameworks we employ in our project.
The second rule: independent
Ideally, we want to set up each test in a fresh environment and tear down all the side effects it introduced to the environment afterward. A side effect can be an altered registry key, an idle process, a dirty table, or browser cache and history. If we neglect to clean up after ourselves, it will make diagnosis more difficult.
The third rule: repeatable
Testing interactions with a user from the outside world is very different from testing functions. The latter has a set of well defined inputs and outputs. On the other hand, behaviors can vary from browser to browser and platform to platform.
This is often the largest hurdle we have to leapt through in our acceptance testing process. It's also the most valuable because we learn cross-platform, cross-browser, and other configuration issues before release.
The last rule: communication
User acceptance testing is a great opportunity to work with our client. We sit down together and write down specifications of the project and turn them into test code that can be easily understood by any stakeholder regardless of their technical background.
Through user acceptance testing we involve clients in production work and are able to adjust to their expectation much faster.
All in all, writing acceptance tests was probably one of the best ways for me to come into a project. A well written suite of acceptance tests is a great source of compiled documentation that can tell you exactly what the application does and what the client wants.
